engineering-decisions

Pass

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and analyze external project data, including existing codebases, Architecture Decision Records (ADRs), and design specifications, to determine project conventions and make decisions. This creates a surface where instructions embedded in project files could influence agent behavior.
  • Ingestion points: Sections 2, 8, 9, and 11 instruct the agent to read existing implementations, design documents, and architecture records.
  • Boundary markers: There are no instructions to use specific delimiters or to ignore embedded instructions within the project files being read.
  • Capability inventory: The agent is tasked with reading files and making autonomous design and implementation decisions based on that content.
  • Sanitization: The skill does not specify methods for sanitizing or escaping the content retrieved from project artifacts.
  • [SAFE]: The skill incorporates several security-positive instructions and constraints:
  • Section 11 establishes a checklist for evaluating new dependencies, explicitly requiring checks for maintainer trust, executable behavior, remote communication, and secret requirements.
  • Section 12 mandates respect for trust boundaries and explicitly prohibits the agent from searching for bypasses to policy restrictions or using alternate tools to evade permissions.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 20, 2026, 08:48 PM